HALO AI Publications

Uma Shanmugasundaram, et al, JCI Insight, 2020
Shanmugasundaram and colleagues used a nonhuman primate model to study T-cell responses associated with latent tuberculosis infection (LTBI). LTBI patients are asymptomatic and are thought to have contained the M. tuberculosis bacteria within granulomatous lesions in the lung. This research group wanted to characterize the immune responses associated with LTBI in order to understand how to prevent the progression to active tuberculosis (TB). In this JCI Insight publication, researchers report that rhesus macaques with LTBI have high levels of M. tuberculosis-specific T cells that produce both IFN- and IL-17. These T cells also express both CXCR3 and CCR6. HALO AI was used to identify percentage of lung tissue composed of granulomas. In addition, the density of CD4+CXCR3+ cells in lung tissue of non-human primate was determined with HALO image analysis. The Tissue Classifier Add-on and the Highplex FL module of HALO were used for quantification of CD4+, CXCR3+, CD68+CD163+, and CD4+CXCR3+ from immunofluorescence microscopy images.

Henning Hvid, Mikala Skydsgaard, Nikolai K. Jensen, Birgitte M. Viuff, Henrik E. Jensen, Martin B. Oleksiewicz, Peter H. Kvist Toxicologic Pathology | First Published August

Allan Rasmusson, et al, The American Journal of Pathology, 2020
An international consortium of researchers characterized lung tissue from patients with COVID-19 using transcriptomic, histologic, and cellular analyses. Nienhold and colleagues report two phenotypes associated with lethal COVID-19 disease. One showed high levels of interferon stimulated genes in the lungs as well as limited lung damage and high levels of cytokines and viral loads. The second phenotype included severe lung damage with low levels of interferon stimulated genes, low viral loads, and high levels of CD8+ T cells and macrophages. As patients with the first phenotype die sooner, this highlights the need for biomarkers to classify COVID-19 patients and potentially guide treatment. HALO AI was trained using annotations from a pathologist to identify lung tissue and the resulting output was confirmed by pathology review. HALO was also used for quantification of immunohistochemistry analysis of CD3, CD4, CD8, CD20, CD68, CD123, CD163, and PD1.

Mesothelin (MSLN) is a cell surface glycoprotein present in many cancer types. Its expression is generally associated with an unfavorable prognosis. This study examined the

Hoa Hoang Ngoc Pham, et al, The American Journal of Pathology, 2020
Pham and colleagues set out to address high false positivity of lymph nodes metastasis analysis using deep learning. As characterizing lymph node metastases in breast and lung cancer is of great clinical importance for treatment selection and prognosis, finding a method with high sensitivity and specificity would represent a major advance. Here, the researchers demonstrate a two-step approach with HALO AI where the first deep learning algorithm excludes the lymph germinal centers that are the source of false positivity and the second algorithm detects tumor cells. The researchers demonstrate this method on lung cancer lymph tissue and find a sensitivity ~78% and specificity ~97% and conclude that a two-step approach can successfully be used to detect lung cancer metastases to the lymph nodes with high specificity. Future research may target development of an algorithm or algorithms with increased sensitivity that maintain high specificity.

David R. Martin, et al, Archives of Pathology & Laboratory Medicine, 2019
Researchers from the University of New Mexico set out to investigate tissue classification using deep learning to evaluate nonneoplastic gastric biopsies. Ground truth diagnosis was established by gastrointestinal pathologists. HALO AI was trained to recognize Helicobacter pylori (H pylori) mediated gastritis, chemical gastropathy, normal mucosa, smooth muscle, and glass. The HALO AI classifier showed high sensitivity and specificity for control biopsies and gastropathy cases and represents the first deep learning driven evaluation of inflammatory gastrointestinal pathology published. The sensitivity and specificity was as follows: normal tissue (73.7% and 79.6%), H pylori (95.7% and 100%), and reactive gastropathy (100% and 62.5%). Martin and colleagues conclude that a convolutional neural network such as HALO AI can function as a screening aid for H pylori gastritis.